[PATCH v7 08/16] arm64: ras: Handle memory failure for uncorrectable errors
Ruidong Tian
tianruidong at linux.alibaba.com
Tue Jun 2 00:15:31 PDT 2026
When an uncorrectable error (UE/DE) is detected and the error record
reports a System Physical Address (SPA), invoke memory_failure() to
offline the affected page. This prevents further consumption of
corrupted data.
Signed-off-by: Ruidong Tian <tianruidong at linux.alibaba.com>
---
arch/arm64/include/asm/ras.h | 4 ++++
drivers/acpi/arm64/aest.c | 5 ++++-
drivers/ras/arm64/ras-core.c | 21 +++++++++++++++++++++
drivers/ras/arm64/ras.h | 26 ++++++++++++++++++++++++++
include/linux/acpi_aest.h | 3 +++
5 files changed, 58 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/include/asm/ras.h b/arch/arm64/include/asm/ras.h
index 42900e1e9a19..7bef631a395c 100644
--- a/arch/arm64/include/asm/ras.h
+++ b/arch/arm64/include/asm/ras.h
@@ -31,6 +31,10 @@
#define ERR_STATUS_UET_UEO 2
#define ERR_STATUS_UET_UER 3
+/* ERR<n>ADDR */
+#define ERR_ADDR_AI BIT(61)
+#define ERR_ADDR_PADDR GENMASK_ULL(55, 0)
+
/* ERR<n>CTLR */
#define ERR_CTLR_CFI BIT(8)
#define ERR_CTLR_FI BIT(3)
diff --git a/drivers/acpi/arm64/aest.c b/drivers/acpi/arm64/aest.c
index 5733c91c8e0d..1b020ab7eccd 100644
--- a/drivers/acpi/arm64/aest.c
+++ b/drivers/acpi/arm64/aest.c
@@ -153,6 +153,9 @@ aest_init_node_props(struct acpi_aest_hdr *hdr, struct property_entry *props,
props[(*p)++] = PROPERTY_ENTRY_U64_ARRAY_LEN("arm,status-reporting",
status_reporting,
group_len);
+ props[(*p)++] = PROPERTY_ENTRY_U64_ARRAY_LEN("arm,addressing-mode",
+ addressing_mode,
+ group_len);
props[(*p)++] = PROPERTY_ENTRY_U64("arm,error-group-base",
common->error_group_register_base);
props[(*p)++] = PROPERTY_ENTRY_U64("arm,fault-inject-base",
@@ -173,7 +176,7 @@ aest_init_node_props(struct acpi_aest_hdr *hdr, struct property_entry *props,
static int __init
aest_create_node_fwnode(struct acpi_aest_hdr *hdr, struct platform_device *pdev)
{
- struct property_entry props[15] = { };
+ struct property_entry props[16] = { };
int p = 0;
int ret;
diff --git a/drivers/ras/arm64/ras-core.c b/drivers/ras/arm64/ras-core.c
index 8c6d202882ed..babb390b795f 100644
--- a/drivers/ras/arm64/ras-core.c
+++ b/drivers/ras/arm64/ras-core.c
@@ -127,7 +127,17 @@ static void ras_print(struct ras_record *record, struct ras_ext_regs *regs)
static void ras_do_proc(struct ras_record *record, struct ras_ext_regs *regs)
{
+ u64 status = regs->err_status, addr = regs->err_addr;
+
ras_print(record, regs);
+
+ if (status & ERR_STATUS_CE)
+ return;
+
+ if (record->addressing_mode == AEST_ADDRESS_LA || (addr & ERR_ADDR_AI))
+ return;
+
+ memory_failure_queue(addr & PHYS_MASK, 0);
}
static void ras_panic(struct ras_record *record, struct ras_ext_regs *regs,
@@ -360,7 +370,10 @@ static int ras_init_record(struct ras_record *record, int i, struct ras_node *no
record->access = &ras_access[node->access_type];
record->index = i;
record->node = node;
+ record->addressing_mode = test_bit(i, node->addressing_mode);
+ ras_record_dbg(record, "record initialized, addressing mode: %s\n",
+ record->addressing_mode ? "LA" : "SPA");
return 0;
}
@@ -598,6 +611,11 @@ static struct ras_node *ras_init_node(struct platform_device *pdev)
GFP_KERNEL);
if (!node->status_reporting)
return ERR_PTR(-ENOMEM);
+ node->addressing_mode = devm_bitmap_zalloc(dev,
+ node->group->errgsr_num * BITS_PER_TYPE(u64),
+ GFP_KERNEL);
+ if (!node->addressing_mode)
+ return ERR_PTR(-ENOMEM);
ret = device_property_read_u64_array(dev, "arm,record-implemented",
(u64 *)node->record_implemented,
@@ -605,6 +623,9 @@ static struct ras_node *ras_init_node(struct platform_device *pdev)
ret = ret ?: device_property_read_u64_array(dev, "arm,status-reporting",
(u64 *)node->status_reporting,
node->group->errgsr_num);
+ ret = ret ?: device_property_read_u64_array(dev, "arm,addressing-mode",
+ (u64 *)node->addressing_mode,
+ node->group->errgsr_num);
if (ret)
return ERR_PTR(ret);
diff --git a/drivers/ras/arm64/ras.h b/drivers/ras/arm64/ras.h
index c26a0aae26c5..11c6def1e4bf 100644
--- a/drivers/ras/arm64/ras.h
+++ b/drivers/ras/arm64/ras.h
@@ -70,6 +70,16 @@ struct ras_record {
const struct ras_access *access;
int index;
+ /*
+ * This bit specifies the addressing mode to populate the ERR_ADDR
+ * register:
+ * 0b: Error record reports System Physical Addresses (SPA) in
+ * the ERR_ADDR register.
+ * 1b: Error record reports error node-specific Logical Addresses (LA)
+ * in the ERR_ADDR register. OS must use other means to translate
+ * the reported LA into SPA.
+ */
+ int addressing_mode;
};
struct ras_group {
@@ -116,6 +126,22 @@ struct ras_node {
* error events.
*/
unsigned long *status_reporting;
+ /*
+ * This bitmap specifies the addressing mode used by each
+ * error record within this error node to populate the
+ * ERR<n>_ADDR register.
+ * Bit[n] of this field pertains to error record corresponding
+ * to index n in the error group.
+ * Bit[n] = 0b: Error record at index n reports System
+ * Physical Addresses (SPA) in the ERR<n>_ADDR
+ * register.
+ * Bit[n] = 1b: Error record at index n reports error
+ * node-specific Logical Addresses (LA) in the
+ * ERR<n>_ADDR register.
+ * OS must use other means to translate the reported LA
+ * into SPA
+ */
+ unsigned long *addressing_mode;
struct ras_record *records;
u32 specific_data_size;
diff --git a/include/linux/acpi_aest.h b/include/linux/acpi_aest.h
index 9cb0fcb52c39..9a8aa234d9e5 100644
--- a/include/linux/acpi_aest.h
+++ b/include/linux/acpi_aest.h
@@ -13,6 +13,9 @@
#define ACPI_AEST_PROC_FLAG_GLOBAL BIT(0)
#define ACPI_AEST_PROC_FLAG_SHARED BIT(1)
+#define AEST_ADDRESS_SPA 0
+#define AEST_ADDRESS_LA 1
+
/* AEST interrupt */
#define AEST_INTERRUPT_MODE BIT(0)
--
2.51.2.612.gdc70283dfc
More information about the linux-arm-kernel
mailing list